CHALLENGES OF SEEING UNDERWATER – A VISION FORTOMORROW

John R. PotterAcoustic Research Laboratory, EE Department & Tropical Marine Science Inst., NUS

http://arl.nus.edu.sg

Abstract

The last decade of this millennium is experiencing a revolution in underseatechnologies relevant to naval military applications. Ten years ago the single mostpressing global consideration of western navies was the tracking and containment ofSoviet nuclear-powered submarines in deep water. These submarines were relativelynoisy by modern standards, and could usually be detected and tracked at considerablerange by passive sonar at low frequencies. The US spent literally billions of dollarson the SOSUS system to implement this strategy. Now that the cold war is over, theemphasis has shifted to higher frequencies, broader bandwidths, intelligent processingand active systems as the focus has moved to littoral regions and submarines havebecome very much quieter. Extremely quiet conventional (even anaerobic)submarines are available from several European sources, some of which are notnecessarily as discriminate as the US and its partners would like about theircustomers in the face of pressing need for foreign exchange. Such submarinesthreaten even the largest and most protected capital groups at sea. At the other end ofthe spectrum, the ease and low cost with which a relatively unsophisticated mine canbe laid in shallow water belies the enormous damage that it can inflict. Opposingthese worrying trends, the maturation of sophisticated technologies in MHz frequencyand parametric sonar, acoustic signal processing, ambient noise imaging, autonomousvehicles, remote sensing, and many other component technologies has opened up arich panoply of options for the modern navy to exploit and deploy. The challenge isin the integration of these nascent technologies to give a meaningful and co-ordinatedpicture.

Introduction

The title of this paper begins with the words ‘Challenges of seeing underwater…’.By ‘seeing’, we mean perceiving our environment in any way that permits images tobe formed. In water, light does not travel very far. Sound, however, can travel tensof thousands of kilometres, e.g. see Munk et. al’s book on tomography1 page 337.Underwater acoustics has therefore traditionally been considered the first choice indetecting and classifying objects underwater, although sonar was rarely used to formimages as such until recently. Now we are familiar with the impressive 3-D plots ofthe seabed obtained from sidescan sonar, synthetic aperture sonar (SAS) andinterferometric multibeam systems. Nevertheless, the displays and types of sonarused for military combat systems still do not enjoy this level of visual sophistication.At the same time, several novel technologies are emerging as promising alternativesor compliments to active and passive sonar. These include Ambient Noise Imaging,Raster scanning lidar (laser imaging), very-high frequency (VHF) active imaging

systems, artificial intelligence signal processing and many more. This paper toucheson a few of these developments and indicates how they might impact naval militaryapplications in the coming decade to provide a vision for the future. Severaltechnology areas will be presented, each in their own section. We begin with themost distant, remote sensing.

Remote Sensing

Remote sensing is familiar to us as a tool to image the earth’s surface. The first sub-surface images to be obtained by satellite were of major features on the ocean seabed,thousands of meters deep, obtained by a superb and highly-accurate set ofmeasurements of average sea surface elevation by the Ocean Dynamics Satellite(Seasat-1) satellite in 1978. Seasat was designed to provide measurements of sea-surface winds, sea-surface temperatures, wave heights, internal waves, atmosphericliquid water content, sea ice features, ocean features, ocean topography, and themarine geoid. Seasat provided 95% global coverage every 36 hours. The instrumentpayload consisted of

• X-band compressed pulse radar altimeter (ALT).• Coherent synthetic aperture radar (SAR)

• Seasat-A scatterometer system (SASS)• Scanning multichannel microwave radiometer (SMMR)• Visible and infrared radiometer (VIRR).

This remarkable instrument brought into focus the enormous range of measurementsand interpretations that could now be made from remote sensing satellites. Theridges and canyons of the sea bed were imprinted on the average sea surface asclearly as a National Geographic map.

One of the more recent satellite remote sensing capabilities to have made an impact inthis regard has been the European synthetic Radar Satellites ERS-1 and ERS-2.

ERS-1 was conceived as an orbiting platform that would be capable of measuring theEarth's atmospheric and surface properties with a high degree of accuracy. It usesactive microwave emissions to collect global measurements and images independentof time of the day or weather conditions. It also undertakes the measurements ofmany parameters not covered by existing satellite systems, including those of seastate, sea surface winds, ocean circulation and sea and ice levels. ERS-1 & 2 carry anumber of instruments consisting of a core set of active microwave sensors supportedby additional, complementary instruments:

• Active Microwave Instrument (AMI), which combines a Synthetic Aperture

Radar (SAR) operating in image or wave mode• Wind scatterometer• Radar Altimeter (RA)

• Along-Track Scanning Radiometer• Microwave Sounder (ATSR)

• Precise Range and Range-rate Equipment (PRARE)

• Laser Retroreflectors (LRR).

Although ERS-1 is now nearing the end of its life, ERS-2 is operating successfully.

With synthetic aperture radar (SAR), very slight variations in the sea surfaceroughness (to which radar wavelengths are very sensitive in terms of the reflectivity)can be detected and mapped. Such variations are caused by the dilation andcompression of the small-scale sea-surface waves, effectively advected by large-scalesurface currents caused by internal waves. As an internal wave propagates, itgenerates tubular convection cells that create small surface currents of alternatecompressive and dilative divergence, resulting in bands of light and dark SARimages.

Internal waves are not the only cause of modifications in the sea surface small-scale

roughness. A ship leaves a very long wake expression on SAR images2. Thepropeller turbulence causes a change in surface roughness that can persist for severalkilometres behind the vessel. Fig. 1 shows two such vessels (bright) and their linearwakes (dark). The wakes are offset from the vessels due to an artefact in SARprocessing which can be used to estimate vessel speed. Researchers at CRISP at

NUS in Singapore are developing automated techniques for detecting and classifyingships from SAR wake images.

The remarkable part of this observation is that the vessel need not be on the seasurface to leave such a trail. SAR images exist that show several naval vesselsproceeding in convoy, each leaving a turbulent wake signature behind. The exciting

thing is that there are two more wakes than visible ships3. Two wakes are notterminated by a bright SAR reflective spot as expected from a surface vessel (whichacts like a very effective corner reflector) but by nothing at all. The conclusion is thatsubmerged vessels left these two turbulent wake trails, invisible by direct means.

Figure 1. ERS-2 SAR image (original copyright ESA 1996) showing two shipsand their turbulent wakes. The wakes are offset from the ships’ headings due to

SAR processing artefacts, from which ship speed can be inferred.Image courtesy of CRISP website at http://www.crisp.nus.edu.sg.

One might expect such turbulent wake fields to be particularly evident in shallowwater, where submerged vessels cannot be far below the surface and the turbulenceenergy is confined by the water depth.

There is now a new French satellite (SPOT) that operates in the visible EM band andwhich can be used to accurately measure the ocean surface colour. In regions of algalbloom and other similar biological activity, or where rivers with high suspendedsediment flow into the sea, one effectively has a colour tracer in the ocean. Asubmerged vessel would stir up the water column and can be expected to leave asurface expression of different colour where such tracers exist.

Coming closer to earth, visual EM systems can be used to directly image dark objectsbelow the sea surface. A hyper-spectral camera with a very high frame rate has beenused with great success to image whales for the purposes of carrying out a population

census4. Although the whales do not appear on the raw colour image, specialenhancement techniques (based on colour correlation differences between surface andsub-surface reflected energy) can be used to subtract the sea-surface reflection andreveal the dark objects beneath. In Fig. 2 the raw hyperspectral data can be seen inthe upper panel, and the surface-reflected corrected data in the lower. A breakingwave event can be seen as a white streak in the upper panel, and appears dark in the

Figure 2. Upper image is raw hyperspectral data of the sea surface. Lowerimage is after correction to remove surface reflection. Two submerged whalesare observed in the lower right area, indicated by the red pen. Figure adapted

from4

lower. Two whales, invisible in the raw data, are clearly delineated in the processeddata of the lower panel, circled in red pen.

The use of laser airborne bathymetry systems, again from aircraft, has also provenvery effective for surveying depths under 50m. An Australian system has now beenextensively tested and is in productive service, surveying difficult areas far more

quickly and effectively than a surface vessel ever could5. An example image fromairborne LIDAR bathymetric mapping adapted from this paper is given in Fig. 3. Thehigh resolution (1.5m) and detail in the estuary depth contours far exceed that of

conventional charts, suggesting that the same type of system, obviously withmodified processing, could perhaps be used to detect submerged vessels andsuspicious features on the sea bed in relatively shallow water.

Autonomous vehicles and probes

At the highest end of this group, and perhaps still strictly in the remote sensingcategory, are the autonomous automated aircraft drones. These can fly missions thatcan carry the type of hyperspectral cameras and radars that we have seen produceuseful information when deployed from satellites and conventional aircraft. In Navysupport, these drones would be capable of surveillance duties over large areas withhigh resolution at a fraction of the cost of piloted aircraft.

Figure 3. Airborne LIDAR bathymetric mapping of an estuary. The high-resolution (1.5 m) and consistent accuracy suggests that airborne LIDAR could be

useful in shallow waters for the detection of submerged submarines.

Dipping into the water itself, autonomous underwater vehicles (AUV’s) have nowmatured, and many vehicles are now available for routine operational surveying tasks.

Although government agencies still support much of the new AUV development,purely commercial companies are now beginning to produce AUV’s as operationalproducts for direct sale, such as the ‘OKPO’ from Daewoo Heavy Industriesillustrated in Fig. 4. Combined with ‘smart’ mission profile interpretations, collisionavoidance sonar, underwater acoustic modems and accurate inertial guidancepackages, these AUV’s are now able to carry out sophisticated surveillance tasks,patrolling areas to seek and map intruders and unfamiliar objects. A small fleet ofAUV’s can carry out patrolling and mine-detection duties over a far larger area than asingle surface vessel towing a remotely-operated vehicle (ROV) and with far bettercoverage. The US is already using a covert AUV equipped with a periscope topenetrate hostile waters and map minefields.

In addition to the traditional, large AUV (which is still expensive, though much lessso than a small surface vessel) there are now various autonomous probes appearingon the market that can carry out profiling tasks and report their findings for RapidEnvironmental Assessment (REA) from a variety of support platforms. Some ofthese probes are expendable deployed from ships or aircraft, such as the shallowwater expendable environmental profiler (SWEEP)6. Others can be recovered a fewhours or days after deployment, such as Ocean Sensors’ Autonomous ProfilingVehicle (APV) or the moored portable ambient noise data acquisition (PANDA)system developed by the ARL and PORL of the Tropical Marine Science Institute(TMSI) in NUS, Singapore. The use of such probes drastically widens the scope for

Figure 4. Daewoo Heavy Industries ‘OKPO’ Commercial AUV.(Courtesy of Daewoo Heavy Industries Ltd)

multi-platform information fusion and exchange, improved systems performanceestimation and tactical decision making. The ability to absorb multi-dimensionalinformation effectively brings us to our next topic, data fusion.

Underwater laser scanning systems

Underwater laser raster scanning imaging systems have now successfully transitionedfrom prototypes to operational commercial products. Several companies offercommercial products, and at least one research institute in the US (Harbour BranchOceanographic) has a specialised group that routinely builds laser scanning imagingsystems to specification. Four of these systems are described in Table I, adapted

from a paper by an HBOI researcher and colleagues7.

NCEL USM HBOI DukeApplication ROV

inspectionBiologicalmapping

Biologicalmapping

Geologicalmapping

Range 0.5-2.0 m 10 cm 0.5 m 2.0–3.0 mAngular Field(air)

40º x 40º 28º x 28º 28º x 28º 30º x 30º

Angular Field(water)

30º x 30º 21º x 21º 21º x 21º 22º x 22º

Accuracy 1 in 1000 1 in 200 1in 200 1 in 100PixelResolution

1342 to 8002 20022 2002 1002

Scan Rate < 4 s 0.05 s 0.05 s 0.20 sLaser Power 150 mW 10 mW 10 mW 75 mWLaserWavelength

532 nm 670 nm 670 nm 532 nm

Depth Rating 914 m 914 m 914 m 4,500 mViewports Acrylic, dome Acrylic, flat Acrylic, flat Acrylic, flatHousing Size 57”L x 9” OD 28”L x 8” OD 29”L x 8” OD 12”&20”L x

9”1 ODHousingMaterial

Anodized AL Anodized AL Anodized AL Titanium

Table 1. Design specifications for laser line scanners developed by HBOI (fromKocak et. al, 19997)

As Table I illustrates, the HBOI systems are limited in range to a few metres.Commercial systems are available which can penetrate up to 10-20 m in most waters,but in turbid water the effective range may be considerably less. Clearly, laser-scanning imaging can provide excellent resolution for close-up classification, butcannot yet fulfill the need for detection at any but the closest ranges.

Recently, several laboratories have demonstrated the use of temporal modulation andsubsequent synchronous laser detection to improve imaging of subsurface objects inshallow water environments8. The technique relies on the synchronous detection ofthe reflected signal, a principle well known in signal detection theory. The advantageof this approach is that the signal return from the target is coherent with thetransmitted waveform while the return produced from scattering is quasicoherent andrepresents noise. The idea is to achieve “processing gain’ via correlation of the

transmitted and received signals against a “noise” background. If this approach canbe made robust, it will extend the range of current lidar systems.

Active sonar

Active sonar is, of course, the oldest subsurface naval technology for detecting andclassifying targets. This is not to say that progress and new developments haveceased. With the shift from passive to active detection of submarines following theend of the cold war and the rapid fall in radiated noise levels from (especially smallnon-nuclear) submarines, there has been a flurry of new developments. Three of themajor new areas are in Low Frequency Active Sonar (LFAS), Parametric sonar andvery high frequency imaging systems.

LFAS is a concept promoted by the US Navy to detect submarines at ranges largeenough to prevent unsuspected attack. It is particularly effective in the protection of acapital battle group in deep water. I would have like to have shown a picture ordrawing of LFAS in this paper, but such images are classified. The principle is todetect a submarine while still outside torpedo range, and then to scare it away or toneutralise it by dispatching a destroyer or similar vessel to deal with the threat. Thesonar relies on the ability of the deep ocean to propagate low frequency sound withlittle loss. An ideal frequency band for propagation might be in the range of 20-100Hz. Unfortunately, small submarines, perhaps of 6m hull diameter, have a low back-scattering cross-section at these frequencies due to diffraction. An ideal detectionfrequency might therefore be somewhat higher, closer to 200-500 Hz. The US LFASsystem uses a number of signals spanning the range 100-500 Hz, including windowedCW and various FM chirps. Details are classified. Again, different signalmodulations optimise different aspects of the detection and classification system.Hyperbolic chirps are excellent for range resolution. Windowed CW is good forDoppler speed estimates. Longer signals optimise processing signal to noise ratio byusing pulse compression. The length of a signal must be optimised for the temporalcoherence of the situation, which will change with location and target geometry.

In practice, the choice of a signal or signals from a library of options is madeadaptively depending on the findings of previous signals, the local environment andthe need to detect, or classify a pre-detected target. LFAS embodies several of theemerging technological features to be found in most leading-edge systems

• Broadband• Adaptive• Bi- or Multi-static

• ‘Smart’ processing

Gone are the days of designing monostatic proprietary systems with fixed transmitand receive characteristics. Modern systems are open-architecture, broad bandwidthand adaptive in terms of the way they can be operated at all levels. This includesevery aspect from the transmitted signal to the processing of the received echo. Suchsystems are now routinely constructed from Commercial Off-The-Shelf (COTS)components. This both reduces the cost, increases the reliability and maintenance

options and allows for more flexible growth and upgrading in the future developmentof the system.

The second major area of active sonar development is in the use of parametric sonar.While the LFAS system is tailored for deep water, parametric sonar is findingadvocates in the area of shallow-water active sonar. The biggest problem facingactive sonar in shallow water is reverberation. Shallow-water active systems are notusually limited by ambient or system noise. The difficulty is that the bottom andsurface interaction is so strong that backscattering from these surfaces dominates thereturn echo. This effect scales with output power, so cannot be easily overcome. Oneway to fight this reverberation is to transmit the signal in a carefully controlled beamwith little or no spatial sidelobes. Traditional beamforming requires large aperturesto achieve low sidelobes, and this results in large and unwieldy transducer arrays.Parametric sonar relies on the non-linear interaction of two high-frequency signals toproduce a lower-frequency beat-frequency emission. The sidelobes of the beat-frequency do not exceed those of the (much higher-frequency) transmit signals.Because sidelobes decrease as the aperture of the transducer array increases (in termsof wavelengths) the result is that a nearly-sidelobe-free transmit beam can beachieved from a compact transducer array operating at much higher frequencies thanthe beat-frequency emission. Relaxation processes quickly absorb high frequencies,so that in the far field, only the (desired) beat frequency remains.

Using parametric sonar, highly collimated beams can be directed into the bottom andsub-bottom to detect buried or partially-buried mines. Parametric beams can also beused to search the water column, steering the beam like a narrow torchlight throughthe region of interest without incurring blinding backscatter from energy which leaksinto angles which interact strongly with the surface or bottom.

The reason that parametric sonars have not been developed for active-duty systemsuntil recently is that the efficiency is extremely low, yielding perhaps only 2-4% ofthe input energy, due to the weakness of the non-linear interaction. Recentdevelopments have stabilised the performance and achieved acceptable levels ofefficiency. A major US sonar company has now built a parametric sonar for the USNavy Fleet. While prototype testing has revealed some shortcomings, the problemsare expected to be resolved in the next year or two and we can expect the littoral USNavy to be parametric-equipped in perhaps five years from now. Another major USsonar company has built several units of a 2-D parametric sonar array for the GermanNavy, where it is in successful operation.

The third area of recent development is in very high frequency (VHF) imagingsystems. Finding objects in turbid waters with underwater video is a problem due tothe lack of visibility. High frequency acoustic signals suffer less scattering in turbidwaters, so attempts have been made to produce an innovation in high-resolutionunderwater acoustic imaging. Until recently, it was not possible to handle thecomputational load associated with sonars with many (perhaps 100-1000) receivingelements at high frequency (100 kHz – several MHz). The new systems coming ontothe marketplace now use sparse arrays, to reduce the number of elements, and modernDSP power to achieve real-time imaging. There are at least two systems of interestthat are now newly available, or in testing and about to become available.

The first system we shall present is a tri-band imaging sonar from a Norwegiancompany that operates at 150, 300 and 600 kHz. The range resolution varies from 50

mm to 100 mm, and the –3 dB beamwidthvaries from 0.6 degrees to 2.5 degrees,with a field of view from 25 degrees to 90degrees, depending on the band9. Thesystem has sufficient computationalpower to run at between 5-10 frames persecond with 1600 receivers, and canimage objects up to 100m range. Thedata rate is some 160 Mbits/s, a largenumber, but still only 10% ofROMANIS’s data bandwidth, which wewill discuss later under ‘Ambient NoiseImaging’. The advantage of this system isthat it can easily be mounted on an ROVto provide a ‘vision’ capacity in turbidwaters easily capable of mine detection,pipeline following, inspection, etc.Furthermore, the system is alreadycommercially available at a competitivecost in the region of S$500k.

The second VHF active system we shallmention is an Australian single-frequencyimaging system, built in conjunction withThomson Marconi Sonar on very similarlines to the Norwegian device butoperating at a frequency of about 3-4MHz10. A prototype system, primarily

intended for acoustic mine imaging, has been successfully tested in turbid waters andis designed to detect the difference in the slot of a straight and Phillips’ head screw.The resolution at close range is of the order of 1 mm. The receiver uses a randomarray of sparse receiving pillars, assembled in collections in ‘smart tiles’ withamplification electronics bonded directly to the ceramic sensors. Although acommercial product is not expected to be available until 2001, one can safely assumethat this device will have a somewhat shorter range than the Norwegian system, in theregion of 5-20 m, but with higher resolution. Such a system would be capable ofhighly discriminate classification and inspection. Operated in tandem with othersensor systems, it could provide the closest-look highest-resolution step of ahierarchy of detection and classification processes.

Ambient Noise Imaging

In 1985 Prof. Michael Buckingham began to develop the idea of Acoustic Daylight(AD) imaging, using ambient noise in the ocean to form images of silent objectsmuch as we use the random light around us to see objects that do not glow in thedark. In 1992 Buckingham and Potter, then at the Scripps Institution ofOceanography, obtained a US Navy research grant to develop a prototype AD

Figure 5. the ‘Echoscope’ VHF activeimaging sonar. Image courtesy ofOmnitech (http://www.omnitech.no)

system. In 1994 this prototype, named the Acoustic Daylight Ocean Noise ImagingSystem (ADONIS), was deployed for the first time. ADONIS was immediatelysuccessful, producing identifiable images with a resolution of 1m at up to 40m range

using only ambient noise11. The results appeared in an award-winning paper in

Scientific American in 199612.

Nevertheless, the images were highly variable in quality, and the statistical vagariesof ambient noise illumination resulted in many false images and unsteady detection.AD was not yet ready to transition to a prototype system. After leaving Scripps, Prof.Potter worked on the signal processing aspects of using ambient noise to image silentobjects with support from the Singapore Defence Research Directorate (DRD). Theresult was the development of several new imaging approaches and a Kalman filtersolution to the problem of how to track targets reliably through varying noiseillumination. This collection of imaging techniques, which include the original ADprinciple but extend far beyond those simple beginnings, has been termed AmbientNoise Imaging (ANI)13.

Although ANI is not yet ready for system implementation, a second-generationresearch ANI imaging array is being built at the Acoustic Research Laboratory (ARL,http:/arl.nus.edu.sg) in Singapore with DRD support. This array, called the RemotelyOperated Mobile Ambient Noise Imaging System (ROMANIS) will be ready at the

end of 1999 and will be tested in local waters in the first half of the year 200014. TheROMANIS system operates in the bandwidth of 25 – 85 kHz, much lower than theVHF imaging systems described above and has an aperture of 1.4m, larger than theVHF systems to retain good resolution. ROMANIS’ angular resolution is similar to

Figure 6. The ROMANIS 25-85 kHz ANI array (left) shrunk to the size of a

Singapore $1 coin (right) for comparison of the form factors.The ROMANIS array is 1.4m in diameter and some 85 mm thick.

the Norwegian imaging sonar, achieved by means of the larger aperture. The datadate from ROMANIS will be some 1.6 Gbit/s, 10 times that of the Norwegianimaging sonar, or any other sonar known to this author. The imaging range will alsobe several hundred m, up to perhaps 1 km, 10 times that of the Norwegian system.The development of ROMANIS is therefore of a technological sophistication on a paror exceeding anything worldwide. The electronic and mechanical sophistication ofROMANIS has resulted in some 500 sensors and more than 50 Pentium PCcomputers being connected without wiring harness and in a thin circular array with aform factor no thicker than a coin, see Fig. 6. This has become possible by theadoption of leading-edge networking protocol technologies (Fibre-channel). Again,the use of commercial off-the-shelf (COTS) technology is being used to streamlinethe design and keep costs down, while improving upgradability and growth paths.Because ROMANIS is fully software configured, it embodies the trend to adaptiveconfiguration, in addition to being broad-band.

If ROMANIS performs as simulations and theoretical calculations predict, it will adda completely covert capability to the current line-up of underwater imaging systems,covering ranges from a few metres to perhaps a kilometre for the detection of largertargets. This capability will be more advanced than anything available outsideSingapore, including the USA. If used as the passive receiver of an active system,ROMANIS’ performance will compliment the range and resolution of VHF activesystems and fill the gap between these and traditional sonars.

While the range limitations of ANI are severe compared to active sonar, theoverriding advantage is that ANI is completely covert and operates againstcompletely silent targets. This cannot be achieved with any other sonar system. Onlylaser scanning systems can compete in this way, and their range is even shorter,perhaps 10-20 m in local turbid waters. In general, a rough rule of thumb is that lasersystems can operate at 2-4 times the range of optical vision. ANI thus offers acomplimentary capability, suitable for the covert investigation of smart mines (whichdetect active sonar interrogations) and other intelligent targets. ANI could be used,for example, to police and patrol restricted areas without alerting intruders that theyhave been detected.

Covert active sonar

The division of sonar systems into active, passive, and now ANI is in fact ratherartificial. As a solution to the limited range of ANI systems, Prof. Potter hassuggested that a hybrid system could be developed, which we shall term covertactive. The idea is to use a broadband passive or ANI receiving system and add anactive capability that transmits signals, which are not readily distinguished fromnatural ambient noise. Potential targets would then not be aware that they were beingilluminated.

These pseudo-random signals could be modelled after snapping shrimp or breakingsurface waves (when high-frequency, broadband transients are useful) or more likeshipping noise (when near-CW would be useful). Broadband high frequency is bestfor robust resolution in target classification. Lower frequency CW is best for long-range detection and Doppler estimation for target speed estimation.

The potential advantages of covert active are enormous. The transmit signals couldbe at very low power, substantially below the true ambient noise and thusundetectable, because the signal would be orthogonal and designed to achieve a highsignal to noise processing gain by the use of pulse-compression when processed witha matched filter. Covert active detection ranges would be similar to active systemsusing similar frequency bands, without the disadvantage of being overt.

The ability to deploy a covert active system requires several advanced technologycomponents, which are now coming online:

• A sufficient understanding of the local ambient noise statistical structure in space

and time. This is required so that the pseudo-random signal can be effectivelycamouflaged. The spatial and temporal structural information can be acquired for‘home waters’ by traditional long-term observation and for novel theatre areas by

using REA techniques including ROV’s and AUV’s as discussed above.• An ability to operate at least bistatic active and preferably multi-static active

sonar topologies. Multi-static active systems require advanced processing and

data fusion combined with sophisticated data link and Communication, Commandand Control (C3) protocols.

• A flexible, software-configured broad-band transmit and receive capability, so

that a wide range of signals can be used from a library covering all frequenciesand signal types of value. Such a capability is a very new feature of sonarsystems, and not yet fully available in most off-the-shelf packages.

Magnetic & electric field detectors & gravimetric techniques

The use of magnetic, electric and gravimetric field sensors to detect objectsunderwater has undergone a periodic ‘pendulum swing’ of interest over severaldecades. The problem with these approaches is generally that the signal is very weakand superimposed on a very high level of noise which is statistically highly variable,so that the natural environment often appears as a target signal, and targets often passunnoticed.

Recent advances in the interpretation of the time-varying nature of both natural noiseand target signatures has resulted in some gains in this regard. Additionally, helium-cooled sensors are now able to detect objects the size and mass of a paper clip at 2mrange. Unfortunately, the sensitivity falls off rapidly with increasing range, so that itrequires a target the size of an aircraft carrier to result in detection at a range of morethan a few nautical miles.

Finally, the improved neutralisation of magnetic signatures from vessels and the useof non-metallic materials moves to negate the advances made in sensor technologyand signal processing.

Data, system and platform fusion

Having briefly reviewed the maturing technologies now being brought to bear onnaval combat, a number of obvious trends become apparent in future trends:

• A greater diversity in types of information (not just sonar and radar)

• A greater number of complimentary systems, working in concert• A greater diversity in types of platforms (satellites, ships, aircraft, ROV’s AUV’s)• A greater number of platforms

The information gathered from all these resources must be assimilated effectively if itis to provide improved tactical decision support. This assimilation process iscomplicated by two main factors:

• The source information is of many different types, spatial and temporal resolution• The source information comes from many different platforms at different rates

The degree of difficulty increases substantially with multi-static sonar systemsreplacing more conventional monostatic or bistatic. The ability to process these dataeffectively and to fuse the result into a single tactical display is a major area of activeresearch, going beyond but including C3. For example, 3-D and ‘CAVE’ (fullyimmersive computational environments) are being explored at several advancedinstitutes, including the IHPC in Singapore, for their potential to display multi-dimensional data.

However, the task goes deeper than these issues. Effective data assimilation andfusion is but the first step. To derive the full benefit of the computational and dataresources, the following assets also need to be seamlessly integrated:

• Historical databases in the form of GIS or similar architecture.• Real-time spot-measurement updates.• 3-D dynamic modelling of circulation and related ocean processes.

The component technologies named above are now in place. The coming decade willsee the effective integration of these assets to provide a truly powerful data fusioncapability, the output of which will be a dynamic and interactive, adaptive tacticaldisplay, available and custom tailored to all commanders in a group, giving ascomplete as possible a tactical picture of the situation and system performances.Research has already yielded encouraging results in using AUV’s to survey andimport data into GIS systems15. The use of minehunting sonar to gather

environmental information for inclusion in GIS has also been explored16.

Signal processing developments

All of the previous sections have dealt primarily with the hardware side of newtechnologies. While many of there require and rely on the high computational powerof modern processors, they do not include the development of new signal processing

schemes and approaches. The last decade has seen a number of very promisingdevelopments and applications in this regard, including:

• Wavelet and other non-Fourier transform techniques• AI and intelligent systems processing

• Statistical and Random Processes estimation in shallow water• Broad band processing• Synthetic aperture and interferometric active sonar

• Shallow-water matched field and inversion algorithms

Wavelet and other transforms have proven to be extremely powerful in the extractionof critical features of many types of signals. Additionally, it is possible to selectwavelet bases which maximise the inter-group separations while minimising theintra-group variability, so that a discrimination algorithm classifies suspect signalswhich a far greater robustness. A high degree of compression of the original data isachieved as an added bonus, permitting results to be linked via a number ofcommunication channels, including acoustic modem. The ARL in the NationalUniversity of Singapore has been very active in this area of signal processing

research, and the results have been most encouraging17,18.

Artificial Intelligence (AI) methods have also been increasingly applied tounderwater signal processing tasks, and there is a large body of literature that hasbeen developed over the last two decades on this area. Coupled with powerfulfeature-extraction transforms, such as the wavelet family, AI is proving to be a veryeffective tool in tactical interpretation and decision-making. The ARL in Singaporeis actively engaged in developing AI and ‘smart’ processing algorithms for imagingand decision-making.

The statistical characterisation of shallow-water environments is difficult, because ofthe high degree of spatial and temporal structure, but necessary for successfuloperation in littoral waters and the prediction of system performance. For obviousreasons, a deterministic description is not possible for many aspects of the operationalenvironment. An understanding of statistical characterisation and estimation istherefore of great importance.

Regrettably, space does not permit further discussion of the many other interestingsignal processing and propagation advances that have been made in the last decadeand which will impact the next.

Conclusion

This decade has seen a powerful shift of interest from deep water, large distant targetdetection and classification to littoral waters, small and varied target detection andclassification at closer range. Not only has this resulted in a predictable shift upwardin sonar frequencies, but it has also brought to play a host of other technologies bywhich we may image the underwater environment. Additionally, signal processinghas moved ahead in the intelligent analysis of signals and automatic classification andprediction. Computational power has continued to double every 14 months or so, and

propagation models have become more complete and accurate, including many moreaspects of the environment and possible mismatch. The challenge is not inidentifying useful technologies, but in bringing them to bear effectively on the tacticalproblems of naval combat scenarios. These technologies will make their mark whenintegrated using the following features:

• Data will be combined from multiple data collection platforms.• Diverse platform types (surface and submarine vessels, ROV’s, AUV’s, drones,

aircraft and satellites) will be used in concert.• Diverse sensing technologies (multi-static passive & active, ANI, covert active,

VHF active, laser) will be employed in a co-ordinated search over all usefulranges, bandwidths, resolutions and degrees of covertness required to spaninterests and needs.

• Systems will be open-architecture and modular to facilitate growth andreconfiguration.

• Systems will employ ‘smart’ processing and AI in the detection, classification and

prediction of environmental developments.• Data fusion and visualisation will be required together with integration of

historical data from GIS and similar sources, real-time spot measurement updates

and 3-D dynamic modelling.

The future thus holds a rich potential for the application of advanced sensortechnologies and signal processing that is already proven and ready to be integratedinto systems. The challenge will be in the fusion.

References

[1] Walter Munk, Peter Worcester, and Carl Wunsch, Ocean AcousticTomography, 1 ed. (Cambridge University Press, Cambridge, 1995).

[2] Christian Melsheimer, “Ship Wake characteristics in SAR and SPOT images,”presented at the Seminar on remote sensing of the marine and coastalenvironment, NUS, Singapore, 1999 (unpublished).

[3] Antony Liu personal communication, 1998.[4] Christina Barnes, Gary Gilbert, Jon Schoonmaker et al., “The possibility of

passive whales tracking with the use of an hyperspectral sensor,” presented atOceans 99, Seattle, Washington, 1999 (unpublished).

[5] Rebecca A. Smith and Geraint R. West, “Airborne LIDAR: a surveying tool forthe new millenium,” presented at Oceans 99, Seattle, Washington, 1999(unpublished).

[6] Robert Tyce, Federico de Strobel, Vittorio Grandi et al., “Shallow waterexpendable and trawler safe environmental profilers,” presented at Oceans '99,Seattle, Washington, 1999 (unpublished).

[7] Donna M. Kocak, Fr ank M. Caimi, Partha S. Das et al., “A 3-D laser linescanner for outcrop scale studies of seafloor features,” presented at Oceans '99,Seattle, Washington, 1999 (unpublished).

[8] V. M. Mullen, M. Strand, and B. Coles, “Modulated Laser Line Scanner forEnhanced Underwater Imaging,” presented at the SPIE, Denver, 1999(unpublished).

[9] P.A.Andersen R.K.Hansen, “A 3D underwater Acoustic Camera - Propertiesand Applications,” in Acoustical Imaging, edited by L.Masotti P.Tortoli(Plenum Press, New York, 1996), pp. 607-611.

[10] Ian S.F. Jones, “High Resolution Underwater Acoustic Imaging,” presented atOceans '99, Seattle Washington, 1999 (unpublished).

[11] M.J. Buckingham and John R. Potter, “Acoustic Daylight imaging: vision in theocean,” in GSA Today (1994), Vol. 4, pp. 97-102.

[12] M.J. Buckingham, Potter John R., and C. Epifanio, “Seeing underwater withbackground noise,” Scientific American February 1996, 86-90 (1996).

[13] J.R. & M. Chitre Potter, “ADONIS imaging with a Kalman filter and higher-order statistics,” presented at the Third European Conference on UnderwaterAcoustics, Heraklion, Greece, 1996 (unpublished).

[14] S Badiu, S Constantin, P Venugopalan et al., “A 1.6 Gigabit/second 25-85 kHzAcoustic Imaging Array – Design Aspects,” presented at Oceans '99, Seattle,Washington (unpublished).

[15] Alexandre Delarue, Samuel Smith, and Edgar An, “AUV Data Processing andVisualization using GIS and Internet Techniques,” presented at Oceans '99,Seattle, Washington, 1999 (unpublished).

[16] Stephen C. Lingsch and William C. Lingsch, “Using a minehunting sonar forreal-time environmental characterization,” presented at Oceans '99, Seattle,Washington, 1999 (unpublished).

[17] Eric. Delory and John R. Potter, “Signal processing aspect of signal detectionmasking and noise suppression,” presented at the Acoustics and Vibration Asia'98, Singapore, 1998 (unpublished).

[18] Eric Delory and John R. Potter, “Comparative study of shift-invariantsymmetric wavelets and cosine local discriminant basis in noisy transientsclassification,” presented at the DSO-NUS joint R&D Seminar, Singapore, 1999(unpublished).